Database of Dynamic Signatures Generated 305YOX1SBF HCM1 SFF SWI5CdH1YOX1SBF HCM1 SFF SWI5CdH1 X 7 X 84400441044204430444044500 10 20 30 40 50 60
% parameters with at least one stable FC020406080100#networks(a) (b)Fig. 2.A perturbation study of the network in (a, top). The histogram in (b) counts
the number of nearby networks with a certain percentage of parameters that exhibit
at least one stable FC. The network in (a, bottom) is one of the top 23 networks that
exhibit more than 40% stable FC.
This observation showed that the backbone network is not the network that
remains after cyclin knockout. However, when we include CDH1 in the net-
work, analysis via DSGRN shows that the network in Fig. 2 (a, top) with CDH1
exhibits no stable full cycles (FCs) across all of parameter space, thus recapit-
ulating the cyclin knockout phenotype. We now address the next question on
how the unknown knocked-out cyclins impinge on the network in Fig. 2 (a, top)
in such a way that their inclusion will cause the full network to oscillate.
To address the question what changes to the network will produce robust
stable FC, we take the network in Fig. 2 (a, top) and add parsimonious num-
bers of randomly chosen nodes and/or edges to the network to sample nearby
networks in the space of all networks. We analyzed 4994 such networks, seeking
stable FCs using DSGRN. The resulting histogram is shown in Fig. 2 (b) with the
number of networks plotted against the percentage of parameters that exhibit
at least one stable FC. Most of the networks (all but 814) show no stable FCs
at all. We assume that a high percentage of parameters exhibiting at least one
stable FC is a reasonable proxy for robustness of full cycle oscillations. There-
fore, we hypothesize that the 23 networks exhibiting at least 40% of parameters
with stable FCs are the best networks in this parsimonious neighborhood of the
original network, and are most likely to represent real regulatory mechanisms
in the cell cycle. One of these top 23 networks is shown in Fig. 2 (a, bottom).
Our method suggests that adding nodesX 7 andX 8 with the appropriate edges
depicted in Fig. 2 (a, bottom) will produce a robustly oscillating network. This
suggestion can be used as a guide for experimentalists to find molecular actors
that fulfill these roles. Taking a different point of view, unchecked progression
through cell cycle is one of hallmarks of cancer. A stable cycle oscillation can